Bringing BAW Technology into Volume Production: The ... - TriQuint

Bringing BAW Technology into Volume Production:
The Ten Commandments and the Seven Deadly Sins
ABSTRACT:
This paper reviews the ten most important things
required to bring a thin film resonator BAW process into
volume production. Relationships between performance
parameters are described and some of the obstacles are
outlined.
Keywords: BAW (Bulk Acoustic Wave), SMR (Solidly
Mounted Resonator), FBAR (Film Bulk Acoustic Resonator)
INTRODUCTION:
The list of companies who have succeeded in
commercializing BAW/FBAR is equally long as the list of
those who have failed and/or given up. Within the next few
years both lists will become a lot longer. It is remarkable
that BAW/FBAR used to be a topic of intense R&D mostly
in companies not previously engaged in Surface Acoustic
Wave (SAW) filters. The schematic cross-section of an
FBAR or SMR BAW device looks very similar to thin-film
capacitors or micromachined pressure sensors; things the IC
industry has been doing for two decades. Lithography
requirements for BAW are trivial as compared to the stateof-the-art
in IC manufacturing. The idea of adding BAW to
the technology portfolio of an IC company is an easy sell.
However in the aftermath of this decision it is very difficult
for management to accept that it takes so long to develop a
BAW process and why progress in performance and
improvement of yield is an uphill battle. In contrast to IC
companies the major players in SAW engaged with
BAW/FBAR very reluctantly and very late, probably
because they had a better understanding on how difficult a
task this may be.
Many of the obstacles found during the development of a
BAW process pop up as a total surprise like rocks blocking a
winding road; very often the engineers have to turn around
and revisit things they had considered completed long ago.
Some examples will be given later. It is also a very common
mistake to trust that good modeling and simulation
capabilities will compensate for deficiencies in resonator
performance. The truth is that there is no software that will
allow you to build good filters if your resonator performance
is insufficient. However it should be pointed out that
simulations and theoretical analysis are essential aids to
figure out how to improve the resonator performance and
what process changes are needed. It can take much longer
than anticipated from a first “almost working” sample to a
Robert Aigner, TriQuint Semiconductor, Florida
raigner@tqs.com
mature process because the improvement of one parameter
usually harms other parameters in an unexpected way.
From a manager’s point of view making a BAW filter seem
like a trivial task. I have heard comments like “How hard
can it be? It is just 8 elements! RF-ICs have millions of
transistors!”. True. But may I ask: how accurate is the gain
of a transistor?
THE “TEN COMMANDMENTS” FOR BAW
1. Coupling coefficient
2. Q-values
3. Uniformity
4. Trimming
5. Spurious modes
6. Temperature coefficient
7. Passivation layer
8. Power handling
9. Nonlinearities
10. Packaging
I) COUPLING COEFFICIENT k 2 eff
Without any doubt this parameter is the biggest challenge in
thin-film BAW. It is quite likely that FBAR/BAW would
have emerged ten years earlier if the deposition methods for
piezoelectric films would have been available. Numerous
early activities in BAW failed because the coupling
coefficients k 2 eff were too low and were not reproducible.
The method of choice for thin-film piezolayers is reactive
magnetron sputtering. The dominant material today is
clearly Aluminum Nitride (AlN). The entrepreneur Ken
Lakin succeeded building the first BAW filters in small
volumes in the mid 1990’s for military applications, but it
was not until the late 1990’s that groups at Hewlett Packard
(now Avago) and Siemens (now Infineon) developed sputter
processes yielding sufficient coupling coefficients for
FBAR/BAW filters in handset applications. A definition of
“good enough” will be given below. The know-how and
process for sputtering AlN with perfect c-axis orientation is
a well kept secret by those who have succeeded. From what
is published about AlN growth [1, 2, 3, 4] there is apparently
no consensus about the conditions under which excellent
film quality is achieved. It is also clear that the “best known
process” developed on one vendor’s tool can not be copied
to another vendor’s tool without significant rework.

There is strong indication that the layer on which AlN is
deposited plays an important role. The following parameters
have been reported to influence k 2 eff significantly:
- material and orientation of bottom-electrode [1]
- smoothness of bottom-electrode surface
- chemical surface condition [3]
Material science groups often present AlN results in terms of
XRD (X-Ray Diffraction) peak-width “rocking-curve
FWHM angle”, it is important to note that a small FWHM
angle is a necessary condition for high coupling but it is by
no means a sufficient condition. Even if XRD shows perfect
orientation there can be an amorphous AlN “starting layer”
on the bottom electrode. The X-ray signal of the amorphous
starting layer would be hidden behind the large peaks of the
oriented AlN. Amorphous layers between the electrode and
the piezolayer harm a resonator badly, the effective coupling
coefficient of a 2GHz resonator can degrade from 6.6% to
6.3% as a consequence of an only 30nm thick amorphous
layer. A method to analyze if an amorphous starting layer is
present is HR-TEM (High Resolution Transmission Electron
Microscopy). In HR-TEM it is actually possible to image
atomic lattice planes of AlN, identify the grains and check if
they are correctly aligned. In bad AlN one may also observe
mis-oriented grains of large size which slowly decay the
thicker the film grows. In such a case the starting layer is not
really amorphous but nevertheless it is a dead zone in terms
of coupling and will harm performance badly.
20 nm
Another case in which XRD would fail to detect a growth
problem is the effect of flipped grains in AlN [3]. Flipped
grains will counteract the actuation of their correctly
oriented neighbors and generate “dead” regions (and
potentially strong acoustic losses). 2% flipped grains will
bring down coupling from 6.6% to 6.3%.
Fig. 2 sketches the hexagonal crystal structure of AlN
(Wurtzite, class 6mm) in the two possible c-axis
configurations. Note that in both configurations hexagonal
monolayers of Aluminum and Nitrogen exist and that it is
therefore not sufficient to start out with the “right” material
as a first monolayer. What distinguishes the two
configurations is the density of Al – N bonds from the first
Al monolayer to the Nitrogen atoms above; in the upper
picture the area density of bonds is a factor of three higher
than in the lower picture. In addition it is reasonable to
assume that the nature of the vertical Al – N bonds is
different from the three other bonds (Al is a group III
element, N is a group V element). Without offering scientific
proof for this hypothesis it appears that growing high quality
AlN requires a seeding condition that provides the right
number and type of bonds during the initial seed formation
phase of AlN. Reference [4] discloses amorphous Silicon as
a good candidate to improve AlN quality.
AlN
Fig. 1: HR-TEM image of AlN on an “as deposited” Tungsten bottom electrode as disclosed in [4] and [6], showing misaligned
grains in the interface region.
W

Fig 2a: AlN crystal structure, c-axis pointing upwards
Fig. 2b: AlN crystal structure, c-axis pointing downwards
(same as 2a but up-side-down)
The most reliable and relevant evaluation of k 2 eff is to build a
simple BAW resonator that can be RF-probed. From the Sparameter
data of resonators, the frequencies of resonance
and anti-resonance can be extracted with high accuracy.
Very short leads must be used between the RF-probe pads
and the actual resonator in order to avoid series inductance
which would result in an overestimation of k 2 eff. Wideband
measurements should always be made to confirm that series
inductance is negligible. Resonators with strong spurious
modes can be problematic for k 2 eff extraction purposes as
those may show distorted phase slopes and will introduce
scatter in the determination of the resonance frequencies. It
is highly recommended to fit a BVD model to the
measurement data rather than just detecting the zerocrossings
of the phase.
The value of k 2 eff can be straight forward calculated from the
resonance frequency fs and the anti-resonance frequency fp.
Unfortunately not all groups working in this field use the
same formulas. On a first look the differences are not big but
one should keep in mind that if k 2 eff drops from 6.6% to
6.3% it can lower yield from 80% down to zero. The
following definitions are used by various groups:
k
k
k
k
2
eff
2
eff
2
eff
2
eff
π f
=
2 f
=
=
=
s
p
1
⎛
⎜
π f
tan
⎜
⎝ 2 f
s
p
⎞
⎟
⎠
2
π ⎛ ⎞ ⎛ − ⎞
⎜
f f p f
s
s ⎟ ⋅⎜
⎟
4 ⎜ ⎟ ⎜ ⎟
⎝ f p ⎠ ⎝ f p ⎠
2
π ⎛ f ⎞ p − f s
⋅⎜
⎟
4 ⎜ ⎟
⎝ f p ⎠
2
π ⎛ f p − f s ⎞
⋅ ⎜
⎟
4 ⎝ f s ⎠
fp/fs = 1.0284 k 2 eff
IEEE standard 6.63 %
2 nd order Taylor 6.62 %
1 st order Taylor 6.81%
Optimist’s favorite 7.01%
IEEE definition
2 nd order Taylor approx.
1 st order Taylor approx.
“Optimist’s favorite”
Table 1 shows by how much the four definitions differ for a
good SMR-BAW. fs = 1866 MHz and fp = 1919 MHz.
The IEEE standard definition is fully consistent with the
results derived for a resonator with infinitely thin electrodes
in classical BAW literature [5]. The reason why this
definition is not widely used is the trouble one runs into
when inverting the transcendent function for modeling
purposes. The 2 nd order Taylor series is a very good
approximation and should be used as a standard. However
most groups and tool vendors use the 1 st order Taylor (or
even the “Optimist’s favorite”) because they show higher
values for k 2 eff.
The values achieved in state-of-the-art deposition tools (on a
regular basis and as an average over full wafer area) are
k 2 eff = 6.7% (using IEEE standard definition) for an
SMR-BAW at 2 GHz.
It should be pointed out that it makes no sense to benchmark
this number against published data for mono-crystalline AlN.
The layer stack and most importantly the electrodes used
have a strong influence on coupling and this has nothing to
do with the AlN quality. A properly designed layer stack can
enhance coupling [6], a badly designed stack will degrade
coupling. In addition to this a thin film piezolayer is laterally
clamped and the d33 parameter of a crystal differs from a
thin-film [1].
Despite the fact that ZnO has in theory a slightly higher
coupling coefficient than AlN it has so far not been
demonstrated as a viable alternative to AlN. The other
prominent piezomaterial PZT is an interesting material for
MEMS and low frequency devices as it has very high
coupling along with extremely high dielectric constant. In
the GHz range PZT appears to have too high intrinsic losses
[7]. Moreover the high dielectric constant and low acoustic

velocity would result in extremely small resonators which in
turn would make it very hard to control acoustic behavior.
II) Q-VALUES
It is a fact that the high Q-values achieved with FBAR were
a key advantage over SAW in the frequency range up to 2
GHz at the time FBAR emerged on the market. In order for
FBAR/BAW to remain competitive against SAW in this
frequency range, the Q-values of next generation BAW
products must improve significantly. Q = 2000 is a
reasonable target.
Most groups working in FBAR/BAW use their own
definition of Q-values. While in all cases the Q-valuedefinition
is related to the impedance-phase-slope of a
resonator, there are different ways to smooth and average the
results obtained. Furthermore, the “de-embedding” of
electrical losses can change results significantly.
Spurious modes may be present in the surrounding of fs
and/or fp and it is not a good idea to use local derivatives to
calculate Q because that can lead to huge scatter in the Qvalue
data extracted. A fit function based on a BVD model
can be used to fit the slope before the extraction of Q is done.
In a really good resonator of reasonable size (20 to 150 ohm
impedance level) the phase-slope is steeper at fp because the
series resistances of the RF-probes, pads and leads have no
effect at this frequency. Whenever a resonator shows a bad
phase-slope at fp it is most likely an acoustic loss that causes
this problem. In order to describe a resonator properly one
would have to define 3 different Q-values:
- the “acoustic” Qa-value represented by the acoustic
branch resistance Ra in Fig.3
- Q@fs which includes the series resistance Rs
- Q@fp which includes Rsi.
R s
R a L a C a
C 0
R si
Fig.3 shows the electrical equivalent circuit of a resonator
know as “modified BVD” [8]
In a good resonator Rsi is extremely small or zero and
Qa = Q@fp. The best values reported to date for FBAR are
Q>2000 [9] and for SMR-BAW are Q=2500 [10].
The Q-values of SMR-BAWs were extremely poor until the
discovery was made by Infineon’s BAW group in year 2000
that shear waves generated as a by-product of thickness
expansional vibration will leak out through the bottom
reflector [11]. Up to this point in time reflectors were built
from quarter wavelength thick layers of high- and lowimpedance
materials. After changing this to fix the dominant
loss mechanism by modifying the reflector to work both for
the longitudinal and the shear wave, the Q-values jumped
from below 700 to above 1300 without any other concurrent
changes.
Hunting down the loss mechanism that presently limits Qvalues
in SMR-BAWs is on the agenda of many groups.
Attempts to explain part of the losses by lateral acoustic
leakage or lateral redistribution currents in the electrodes
[12] have been published. However it is not proven at this
point in time that losses of this nature limit the Q-values we
see in experiments. To my knowledge there is no publication
about a method that would consistently predict the
theoretical limit for Q for FBAR or SMR.
III) UNIFORMITY
Uniformity of layer thickness across a wafer is one of the
obvious challenges in FBAR/BAW. Using a Mason-model
approach one can derive the change of resonance frequency
if a layer thickness is off target. Layer thickness uniformity
has been discussed in many publications [13]. There are
many aspects of uniformity that are less obvious and easily
escape people’s attention:
- uniformity of acoustic parameters (velocity, density) is
equally important to thickness-uniformity.
- Uniformity of k 2 eff across the wafer is related to
maximum yield.
- Some types of uniformity problems can be fixed by
trimming while others can’t. Some layers deep down in
the stack do not change frequency by a significant
amount but they are important with regard to acoustic
wave dispersion.
- Uniformity on a small lateral scale is very important. A
change of layer thickness or material parameters within
one resonator can kill resonator performance. Any
unintended change of parameters between the resonators
comprising a filter or between neighboring filters on a
wafer will cause severe yield losses. No trimming
method is able to fix this type of problem.
While thickness uniformity, as it concerns frequency
tolerance is somewhat overemphasized, the severity of the
other aspects is clearly underestimated. At this point it is
worth mentioning that many of the methods used to measure
and map thickness are indirect and need very careful
calibration. For transparent layers the standard tool is an
optical spectrometer/ellipsometer. It measures thickness very
accurately if refractive index and the optical properties of the
layer underneath are well defined. A change in “optical
length” is not necessarily related to “acoustical length”.
Another example is the thickness measurement of metal
layers. The 4-Point-Resistance method is a standard method
in many IC fabs. The result of the measurement is the
“Rsquare” of a metal layer which is inversely proportional to
thickness. The conductivity is a calibration parameter. It is
wrong to assume that conductivity is in any way related to
the acoustic properties of the film. There is no way to
distinguish if a change in Rsquare is related to a thickness
changes or if conductivity has changed. The 4-Point-
Resistance method is not suited to verify the uniformity of
metal layers for BAW. A somewhat better way to measure
thickness is XRF (X-Ray Fluorescence). The calibration
parameter is density and the assumption is that alloy

composition does not change. Thin and heavy layers can be
measured with XRF quite well but it is not possible to
measure multilayer stacks. The king of metrology methods
for FBAR/BAW is Femtosecond Laser-Pulse Sonar. The
tool measures acoustic delay time(s) of layer(s) and directly
extracts the parameters needed for a Mason model. Even if
the calibration parameter “acoustic velocity” is slightly off,
the error will, to a large extent, self-correct when calculating
the frequency. One of the nice features of this method is the
small spot size (10 .. 20 μm) and the capability to measure
multi-layers in one shot.
IV) TRIMMING
A typical PCS duplexer spec requires hitting a frequency
target accurately to ±1 MHz, which is equivalent to ±0.05%
relative error. Accordingly the layer thickness, acoustic
velocity and density of several layers would have to be
accurate to ±0.02% each. There is no hope to achieve such a
high accuracy “as deposited”. As a consequence frequency
trimming is one of the key processes in FBAR/BAW. Runto-run
variations can be corrected by feed-forward control of
deposition thickness. However correction of uniformity
errors requires “localized processing”. Off-the-shelf
trimming tools for BAW and SAW are now available and
they are all based on local Ion-beam etching. A very narrow
Ion-beam is scanned over the wafer surface at a controlled
speed which determines the local removal that occurs [14].
The challenge in trimming is to choose the right strategy.
Trimming for volume production is a balancing act between
accuracy and throughput. The input data for trimming comes
from mapping the thickness and/or frequency. The mapping
grid must satisfy Shannon’s sampling theorem in both
coordinate directions. The size and shape of the ion-beam is
a very important parameter which determines the maximaum
frequency gradient the tool can handle. Small diameter
beams allow very large gradients but tend to have a very low
volume etch rate which results in very long processing times.
The stability of the beam shape over a long time is crucial
for the accuracy the tool will achieve.
The topmost layer from which the Ion-beam is supposed to
removes material must be chosen considering the following
parameters:
- Trimming into heavy layers allows very high
throughput but there are limitations in accuracy. Even at
maximum scanning speed, the frequency shift can be
larger than desired on certain spots on the wafer. The
maximum speed and acceleration of the mechanical
scanning system can accommodate is an important
parameter for such a case.
- Metal layers may grow a native oxide layer in air. The
thickness of this layer is not predictable. In most cases
this native oxide exhibits a different etch rate than the
metal below. This results in strong nonlinear behavior in
the dose versus removal curve. The worst materials to
trim are Al alloys. Al2O3 etches factor 5 slower than the
Al below.
- Dielectric materials are well suited to trim very
accurately. On the other hand, the frequency shift that
can be obtained is very limited because suitable layers
(SiO2 or Si3N4) have a low mass density and low etch
rate. In general dielectric layers show excellent
reproducibility of trimming rate. The drawback of
dielectric layers is that their presence generally degrades
coupling coefficient without contributing to the
conductivity of the electrode. The maximum thickness
of a dielectric layer is determined by how much margin,
in coupling coefficient, a filter has.
V) SPURIOUS MODES
Spurious modes are related to lateral standing waves in a
BAW or FBAR resonator. Only a very tiny amount of
spurious ripple is acceptable in filters with stringent
amplitude ripple and group delay ripple specs. Moreover the
presence of spurious modes usually goes along with
degraded Q-values. There are different ways to fight
spurious modes in FBARs and in SMRs. For SMRs it is
certain that pronounced spurious modes will show up when
the Q-values exceed 1000 (regardless of resonator shape)
and they will be really bad at Q = 1500. At Q

It is possible to further improve TCF in SMRs by increasing
the SiO2 content and by moving the SiO2 closer to the high
stress regions in the stack. SMRs with essentially zero TCF
have been reported [21]. All of these approaches harm k 2 eff
massively and can only be used for filters and resonators
with small fractional bandwidth. In order to have a zero TCF
one would have to accept k 2 eff of an SMR resonator below
4%.
VII) PASSIVATION LAYER
The purpose of a passivation layer is to protect the
resonators from detrimental effects caused by humidity or
corrosive fluids. Whether they are required or not has been a
topic of heated discussions. This question can only be
answered after reviewing the packaging options and
preferences for FBAR/BAW.
The main challenge of passivation layers for BAW-SMRs at
high frequencies is that, for acoustical reasons, one can not
choose layers as thick as in traditional IC processes. While
IC process typically use a combination of 300nm SiO2 with
300nm Si3N4 on top, the maximum thickness for an SMR at
2GHz is less than 100nm (or else the k 2 eff degrades badly). It
is possible to deposit a pin-hole free layer of that thickness.
This layer can also serve as a “trimming layer”.
VIII) POWER HANDLING
FBAR/BAW devices endure higher power levels better than
SAWs mainly because the electrical currents distribute more
evenly. There are no narrow IDT fingers like in SAW which
are prone to electromigration damage. Even though the
minimum feature size of BAW is much larger, the current
densities can be enormous. For a BAW at 32dB transmit
power at the upper passband skirt (worst case scenario) the
following observation has been made. Depending on the
electrode materials used the combined effect of current
density and mechanical stress will cause the electrode
material to migrate and form rough regions on the resonator
surface. The losses of that resonator will increase and so will
the temperature of the resonator. As migration effects follow
an Arrhenius type law with temperature, the damage
accelerates and the resonator will self-destroy within
minutes. The power handling of BAWs is a strong function
of the ambient temperature as suggested by the Arrhenius
law. It is very important to keep the filter chip as cold as
possible, therefore it is necessary to provide a good heat sink.
The key to excellent power handling is to improve the
electro- and stress-migration properties of the weakest
material involved. This is an exercise the SAW vendors have
successfully completed during the last years [22]. Significant
material research activities will be required to find an
optimum solution for BAW.
IX) NON-LINEARITIES
The discovery of nonlinear behavior in BAW and FBAR
was first reported in 2005 [20]. It has long been known that
solids exhibit nonlinear stress-strain relationships at high
stress levels [23]. The binding forces of the atoms in a lattice
are a strongly nonlinear function of distance. This effect is
described by the 3 rd order elastic constants of a material. In
addition to that the piezoelectric constants change as the
crystal deforms. It should not come as a total surprise that
the elastic constants of a material are modulated by strain
generated by large voltage swings at high RF-power levels
or by bias voltages. To my knowledge no complete theory
about the nonlinear behavior of AlN based BAW resonators
has been published.
There are known tricks to improve the linearity of BAW
filters. Cascading of two double sized resonators to replace
one resonator in a filter is one of them, however this is not
possible for all resonators in a filter as it would increase the
size of a BAW by a factor of 4. Other tricks involve partial
compensation of harmonic tones.
X) PACKAGING.
The fact that BAW filters can be processed on Silicon
wafers is a big advantage for packaging. Silicon is much
easier to handle than Lithium Tantalate or other piezo
materials SAW filters are typically made from. Silicon is
inherently stronger and less fragile, moreover it can be
temperature ramped at a high rate which is a significant
advantage over SAW. The thermal expansion coefficient of
Silicon is small and isotropic and the heat conductivity is
excellent.
FBARs and BAWs need a cavity above the top electrode
which was also true for all SAW filters until the recent
introduction of Boundary-Wave-Acoustic Devices [24].
Providing this cavity in a cost efficient way is a key for
commercial success in consumer markets. The method of
choice is to use a Wafer-Level-Packaging (WLP) approach
in which the cavities are created in a batch process. FBAR
[25] and BAW WLPs [26] are in volume production and
details are published. [25] describes a process using waferbonding
with a hermetic seal while [26] uses a polymer
build-up approach which is not hermetic. The non-hermetic
approach is potentially lower in cost but requires perfectly
passivated resonators which do not corrode in a humid
environment. Passivation is much easier to apply to BAW-
SMR than to FBAR because there is no bottom cavity. For
BAW-SMR a pin-hole free protection layer on the top
surface is sufficient.
SUMMARY
The market share that FBAR/BAW will be able to gather is
largely dependent on how well they support RF-integration.
It is a clear trend in the wireless industry that companies
building phones want to source functional blocks and
modules rather than discrete filters. For the phone
manufacturer the filters are invisible; it is the supplier’s
responsibility to make filters and active components work
together smoothly. The determining factor will be whether
SAW or FBAR/BAW will provide a more cost competitive
solution that meets the necessary performance level.

THE “SEVEN DEADLY SINS”
(AND HOW MANAGERS CAN AVOID THEM)
LUXURIA (extravagance)
Choose your target markets and applications wisely.
GULA (gluttony)
Don’t build more capacity than you can fill.
AVARITIA (greed)
Don’t be “penny wise and pound foolish”; don’t try to
save money in the wrong place.
ACEDIA (sloth)
Not applicable. Eningeers don’t have this bad habit.
IRA (wrath)
Engineers do their best to meet your aggressive
schedules; don’t demand miracles.
INVIDIA (envy)
Don’t think your competitors were just lucky.
SUPERBIA (pride)
Know what SAW is capable of doing and what your
competitors are up to.
ACKNOWLEDGEMENTS:
Many thanks to Ben Abbott for discussions on the crystal
structure of AlN and for introducing me to the world of
SAW.
REFERENCES:
[1] M.-A. Dubois, P. Muralt, L. Sagalowicz, “Aluminum Nitride
Thin Films for High Frequency Applications”, Ferroelectrics, 1999,
Vol. 224, pp. 243 – 250
[2] R. S. Naik, R. Reif, J.J. Lutsky, C. G. Sodini, “Low-
Temperature Deposition of Highly Textured Aluminum Nitride by
Direct Current Magnetron Sputtering for Applications in Thin-Film
Resonators”, Journal of The Electrochemical Society, 146 (2) 691 –
696, 1999
[3] J.A. Ruffner, P.G. Clem, B.A. Tuttle, D. Dimos, D.M.Gonzales,
“Effect of substrate composition on the piezoelectric response of
reactively sputtered AlN thin film”, Thin Solid Films 354 (1999)
256-261
[4] US patent, US6878604B2
[5] J.F. Rosenbaum, “Bulk Acoustic Wave Theory and Devices”,
Artech House, Boston, London, 1988
[6] US patent, US6291391B1
[7] Q-X Su, P. Kirby, E. Komuro, M. Imura, Q. Zhang, R.
Whatmore,”Thin-Film Bulk Acoustic Resonators and Filters Using
ZnO and Lead-Zirconium-Titanate Thin Films”, IEEE Transactions
on Microwave Theory and Techniques, Vol. 49, No. 4, April 2001
[8] J.D. LarsonIII, P.D. Bradley, S. Wartenberg, R.C. Ruby,
“Modified Butterworth-Van Dyke circuit for FBAR resonators and
automated measurement system”, IEEE Ultrasonics Symposium
2000, vol.1 page 863-868
[9] R. Ruby, J. D. LarsonIII, R. S. Fazzio, C. Feng, “Performance
Degradation Effects in FBAR Filters and Resonators due to Lamb
Wave Modes”, Proceedings of IEEE Ultrasonics Symposium 2005,
Rotterdam
[10] G. Fattinger, R. Aigner, S. Marksteiner, “Everything you
always wanted to know about BAW”, Workshop proceedings,
APMC2006, Yokohama Dec. 2006
[11] S. Marksteiner, J. Kaitila, G. G. Fattinger, R. Aigner,
„Optimization of Acoustic Mirrors for Solidly Mounted BAW
Resonators”, Proceedings of IEEE Ultrasonics Symposium 2005,
Rotterdam
[12] R. Thalhammer, R. Aigner, “Energy loss mechanisms in
SMR–type BAW devices”, Proceedings of IEEE IMS-MTT-S 2005,
Long Beach
[13] R. Lanz, et.al. “Aluminum-Nitride Manufacturing Solution for
BAW and other MEMS Applications Using a Novel, High-
Uniformity PVD Source”, Proceedings of IEEE Ultrasonics
Symposium 2006, Vancouver
[14] European patent EP1456947B1
[15] G. G. Fattinger, S. Marksteiner, J. Kaitila, and R. Aigner,
„Optimization of Acoustic Dispersion for High Performance Thin
Film BAW Resonators”, Proceedings of IEEE Ultrasonics
Symposium 2005, Rotterdam
[16] A. Link, et.al “Suppression of Spurious Modes in Mirror-Type
Thin Film BAW Resonators Using an Appropriate Shape of the
Active Area”, Proceedings of IEEE Ultrasonics Symposium 2005,
Rotterdam
[17] US patents US6182619 B1 and US6788170 B1
[18] National Institute of Standards (NIST) ceramic database,
http://www.ceramics.nist.gov/srd/summary/emodox00.htm
[19] R. Aigner, “Volume manufacturing of BAW-filters in a
CMOS fab”, Acoustic Wave Device Symposium 2004, Chiba
Japan, March 2004
[20] R. Aigner, N.-H. Huynh, M. Handtmann, S. Marksteiner
“Behavior of BAW devices at high power levels” Proceedings of
IEEE IMS-MTT-S 2005, Long Beach
[21] K.M. Lakin, “A Review of Thin-Film Resonator Technology”,
IEEE microwave magazine, Dec. 2003
[22] O. Nakagawara, et al. “High power durable SAW antenna
duplexers for W-CDMA with epitaxially grown aluminum
electrodes”, IEEE Ultrasonics Symposium 2002
[23] W.P. Mason, “Physical Acoustics, Vol III, part A” chapter 5,
pp 199. Academic Press, New York and London, 1966
[24] H. Kando, et. al. “RF Filter using Boundary Acoustic Wave”,
Proceedings of IEEE Ultrasonics Symposium 2006, Vancouver
[25] K. Wang, M. Frank, P. Bradley, R. Ruby, W. Mueller, „FBAR
Rx filter for handset front-end modules with wafer-levelpackaging“,
Proceedings of IEEE Ultrasonics Symposium 2003,
Honolulu
[26] M. Franosch, K.-G. Oppermann, A. Meckes, W. Nessler, R.
Aigner, “A Wafer-Level-Process using Photo-Epoxy to create Air-
Cavities for Bulk-Acoustic-Wave RF-Filters”, Proceedings of
IMAPS 2004 conference, Long Beach California, Nov. 2004